HarnessMatch
Menu

Coding agent

Kimi Code

Multi-provider CLI with parallel subagents and local web or IDE surfaces.

Status
Active
License
MIT
Evidence
Documented + code-verifiable, 15 sources
Product record checked
2026-07-27
Compare harnessesInspect sources

At a glance

Moonshot AI's MIT-licensed coding harness provides a terminal UI, structured print mode, multi-provider routing, MCP, skills, plugins, hooks, ACP editor integration, a local web service, trace inspection, and parallel foreground or background subagents.

Good choice if

  • Parallel coding work that uses foreground, background, or swarm-style subagents
  • Teams switching among Kimi, Anthropic, OpenAI-compatible, Gemini, Vertex, and local providers
  • One harness shared across a TUI, structured automation, ACP editors, and a local web interface

Check before choosing

  • Local shell and file tools run with host privileges; no first-party OS sandbox is documented
  • Print mode switches to an unattended auto-permission posture; its background Bash and subagent defaults have no practical wall-clock limit unless configured
  • Hooks fail open on errors or timeouts and must not be the sole barrier for high-risk actions
See 3 more considerations
  • AgentSwarm can ramp without a configured concurrency cap, while delegated agents inherit the caller's allowed permission rules
  • The local web service uses loopback and bearer authentication by default, but --dangerous-bypass-auth exposes sessions, files, and shell access to anyone who can reach the port
  • Plugins may add skills and start MCP servers; third-party and custom sources require explicit trust and user-global installs are not project-scoped

Why it qualifies as a coding harness

This confirms category fit, not product quality. Every required criterion links back to first-party evidence.

Qualifies4 of 4 required criteria evidenced

  • Adaptive agent loop

    Documented

    The system repeatedly observes results and chooses the next action instead of following a fixed one-pass graph.

    EvidenceKimi Code documentation

  • Repository tool execution

    Documented

    The system can use tools to inspect and change a repository or its execution environment.

    EvidenceBuilt-in tools and swarms

  • Task-aware context management

    Documented

    The runtime assembles, updates, compacts, retrieves, or persists task-relevant context while work proceeds.

    EvidenceSessions and context

  • Model-independent runtime control

    Documented

    Permissions, budgets, interruption, policy, or stop controls operate outside the model's own text generation.

    EvidenceCommand reference

Membership establishes category fit only. It does not score quality, safety, autonomy, model capability, or benchmark performance. · Read the membership rule.

How it works under the hood

Seven mechanisms mapped from first-party records. These labels describe what the harness provides, not how intelligent its model is.

7/7layers documented
  • Execution & isolationHost processDocumented mechanism, not a performance score.
  • Tooling & integrationsExtensible toolsDocumented mechanism, not a performance score.
  • Context & stateManaged contextDocumented mechanism, not a performance score.
  • Lifecycle & recoverySession resumeDocumented mechanism, not a performance score.
  • ObservabilityStructured tracesDocumented mechanism, not a performance score.
  • VerificationTool-assistedDocumented mechanism, not a performance score.
  • Governance & permissionsApproval promptsDocumented mechanism, not a performance score.

Public code audit

4/5public artifacts present
Security policy
Present at inspected commit
CI workflow
Present at inspected commit
Automated tests
Present at inspected commit
Evaluation assets
Not found
Contributor documentation
Present at inspected commit

The 0.29.2 tree contains 1,256 test-like files and eight workflows with security and contributor documentation. No dedicated coding-harness evaluation suite or immutable comparative result record was found; files named evaluate implement tool-policy logic rather than product benchmarks.

Inspect commit 8a45f10eddbb35c317047e82e567cdb59a220b4f, checked 2026-07-27

Measured configurations

No benchmark run passes the full metadata admission policy for this harness yet. Missing data is not scored as zero.

Benchmark policy and all runs

Capability support

Documented first-class product support, checked against the sources below.

External tools (MCP)
DocumentedProduct-supported MCP integrationSource · checked 2026-07-27The source establishes the mechanism, not its quality or availability in every mode.
Local models
DocumentedLocal or self-hosted model pathSource · checked 2026-07-27The source establishes the mechanism, not its quality or availability in every mode.
Agent parallelism
DocumentedDelegated or parallel agent workflowSource · checked 2026-07-27The source establishes the mechanism, not its quality or availability in every mode.
Runs without an open UI
DocumentedNon-interactive or automation surfaceSource · checked 2026-07-27The source establishes the mechanism, not its quality or availability in every mode.
Browser control
Not documentedNo first-class support established by the current recordAbsence of current documentation is not proof that the capability is impossible.
Isolated execution
Not documentedNo first-class support established by the current recordAbsence of current documentation is not proof that the capability is impossible.
Undo file changes
Not documentedNo first-class support established by the current recordAbsence of current documentation is not proof that the capability is impossible.

Primary evidence

Each capability claim is tied to a first-party record and a verification date.

Product record checked 2026-07-27
Kimi Code 0.29.2 source snapshotImmutable MIT source snapshot for the audited release, including 1,256 test-like files and eight workflows but no dedicated product-evaluation suiteOfficial repositoryKimi Code 0.29.2 releaseStable package release and dated version boundary used for the auditOfficial announcementKimi Code documentationSkills, hooks, parallel subagents, MCP, installation, TUI, and extension pointsOfficial docsProviders and modelsKimi, Anthropic, OpenAI-compatible, Responses API, Gemini, Vertex, and custom provider routingOfficial docsCommand referenceStructured print mode, permission modes, ACP integration, authenticated local web service, session visualizer, session selection, and output formatsOfficial docsConfiguration defaultsManual interactive permissions, default-on anonymous telemetry, background-task behavior, subagent timeouts, and effectively unbounded print-mode ceilingsOfficial docsBuilt-in tools and swarmsAgent and AgentSwarm approval behavior, inherited permissions, background execution, timeout defaults, and configurable swarm concurrencyOfficial docsSessions and contextStructured local Wire event streams, resume, replay, fork, subagent trace history, and export without claiming learned cross-session memoryOfficial docs
View 7 additional sources
Agents, state and recovery1 sources
Data locationsLocal configuration, credentials, session history, logs, skills, plugin state, relocation, and cleanup boundariesOfficial docs
Automation and extensions1 sources
PluginsOfficial and third-party marketplaces, trust labels, Git and ZIP installation, plugin skills, MCP servers, managed copies, and user-global scopeOfficial docs
Enterprise and operations1 sources
Environment controlsData-root relocation, telemetry disablement, background-task lifetime and concurrency limits, model overrides, and diagnostic controlsOfficial docs
Additional first-party evidence4 sources
HooksLifecycle events, pre-tool blocking, observability hooks, timeouts, and explicit fail-open semanticsOfficial docsAgent SkillsProject, user, extra, and built-in skill scopes, automatic or direct invocation, progressive loading, scripts, and cross-agent directoriesOfficial docsAgents and subagentsBuilt-in and custom agents, parallel isolated contexts, background execution, permission inheritance, tool restrictions, and local trace storageOfficial docsModel Context ProtocolProject and user MCP scopes, local process trust boundary, tool permissions, OAuth, approval behavior, and YOLO-mode warningOfficial docs